Skeleton-bridged Mesh Saliency and Mesh Simplification
Wanna Luan, Chengming Liu, Haibo Pang · 2020
This paper presents a local-to-global framework for mesh saliency, which can measure the visual importance in agreement with human perception. Firstly, based on Voronoi diagram, we give a new descriptor which does not only compute the local saliency, but also facilitates the computation of 3D skeleton and global saliency. Then we calculate the global saliency based on skeleton extraction and hierarchical clustering. In addition, an effective denoising method is exploited to guarantee the smoothness of skeletons and clustering accuracy. At last, we put forward a new solution to integrate local and global saliency to obtain a visually better result. Large amounts of models are tested which demonstrate that our results are consistent with human-marked interest points. We also apply the saliency score on mesh simplification to get visually saliency-aware simplification results.